Related Experiment Video
Updated: Aug 4, 2025
06:08
Quantitative Visualization and Detection of Skin Cancer Using Dynamic Thermal Imaging
Published on: May 5, 2011
16.9K
Deep Hyperspherical Clustering for Skin Lesion Medical Image Segmentation
IEEE Journal of Biomedical and Health Informatics
|April 6, 2023
Summary
This study introduces a novel deep hyperspherical clustering (DHC) method for accurate skin lesion segmentation. DHC improves diagnostic precision by handling data uncertainty and reducing reliance on labeled medical images.
Area of Science:
- Medical Imaging
- Computer Vision
- Artificial Intelligence
Background:
- Skin lesion diagnosis using imaging is challenging due to data uncertainty affecting accuracy.
- Existing methods often require extensive labeled data, limiting their applicability.
Purpose of the Study:
- To develop a novel deep hyperspherical clustering (DHC) method for skin lesion medical image segmentation.
- To improve segmentation performance and characterize data uncertainty without labeled data.
Main Methods:
- Combined deep convolutional neural networks with belief functions (TBF).
- Employed SLIC superpixel algorithm for image grouping and an autoencoder for feature extraction.
- Developed a hypersphere loss function for network training and TBF for uncertainty characterization.
Main Results:
- The DHC method effectively characterizes imprecision between skin lesions and non-lesions.
- Experiments on four benchmark datasets showed superior segmentation performance compared to other methods.
- Achieved increased prediction accuracy and the ability to perceive imprecise regions.
Conclusions:
- The proposed DHC method offers a robust approach for skin lesion segmentation, particularly in managing data uncertainty.
- This technique enhances diagnostic accuracy in medical imaging procedures.
- DHC provides a valuable tool for analyzing dermoscopic images with improved precision.
